DocsCommands
FutureX CLI commands
Every command FutureX accepts, every flag it takes, and the files that configure it. One page, no hunting.
All documentation pages
Getting started
Reference
Commands
One binary, one habit: read the repo first, then act. Run these from your project root. If the CLI is not installed yet, start with the quickstart — install, sign in, first change.
futurex help19 of 19 commands
| Command | What it does |
|---|---|
| futurex | Open the interactive REPL. No arguments, no ceremony. |
| futurex ask "<prompt>" | One-shot answer, then exit. |
| futurex agent "<task>" | Agentic loop: the model uses tools. Reads happen automatically, writes are gated. |
| futurex index | Index the repo so answers can be retrieval-augmented. |
| futurex symbols [name] | Query the AST symbol index. |
| futurex imports [file] | Show the import graph, plus any broken imports. |
| futurex init | Create futurex-memory.md if it is not there yet. |
| futurex compact | Bound the size of futurex-memory.md: regenerate it, archive the old one. |
| futurex plan "<task>" | Pre-build plan: goal, value and budget, research, tooling, and a task DAG. |
| futurex build "<task>" | Plan, then execute against the plan. |
| futurex resume | Resume the latest chat session. --list to see them, --session <id> to pick one. |
| futurex skills [request] | List skills, or route a request to the skills that match it. |
| futurex permissions | Per-project consent: show, set <kind> <state>, or reset. |
| futurex verify [--list] | Project diagnostics: typecheck, lint, test, build. |
| futurex test | Run the project test script. |
| futurex login | Sign in with a browser link code. --no-browser prints the code instead of opening a tab. |
| futurex logout | Clear stored credentials. |
| futurex whoami | Show the current backend and auth mode. |
| futurex help | Show help. |
init and compact both act on the project memory file — how that works is on project memory.
Options
Flags go after the command. Anything scoped to particular commands says so in its description.
futurex agent "add a health check" --tier fx-pro --max-steps 12| Flag | What it does |
|---|---|
| --tier <module> | Pick the module: fx-pro, fx-fast, fx-mini, or fx-eco. |
| --effort <level> | off, high, or max — enforced against the module you picked. |
| -r, --reason | Use more reasoning effort (high). |
| --max | Maximum reasoning. Implies fx-pro. |
| --pro | Use the fx-pro module. |
| -m, --model <id> | Override the underlying model id. Advanced. |
| --max-tokens <n> | Cap output tokens. |
| --no-memory | Ignore futurex-memory.md for this run. |
| --no-repo | Do not include the repo card. |
| --no-context | Do not augment the prompt with retrieved repo chunks. |
| --top-k <n> | Chunks retrieved per turn. Default 6. |
| --auto-approve | agent — run mutating tools without prompting. Dangerous; know why you typed it. |
| --json | ask, agent — JSONL events on stdout, exit 0 or 1. |
| --output-format <f> | json or text. Alias of --json. |
| --max-steps <n> | agent — max model and tool rounds. Default 32. |
| --no-verify | agent — skip the automatic typecheck, lint, and test after edits. |
| --value <tier> | plan, build — low, standard, high, or critical. |
| --budget <n> | plan, build — cap the build token budget. |
| --restart | plan, build — ignore checkpoints and re-plan from scratch. |
| -v, --version | Print the version. |
The four modules — Pro, Fast, Light, and Eco — carry their own default effort and output cap. Which to reach for, and what each one costs, is on models.
Settings files
Two files, read in this order:
~/.futurex/settings.json # your machine
<repo>/.futurex/settings.json # this project| Key | What it sets |
|---|---|
| permissions.allow / permissions.deny | Standing consent rules — the same ground futurex permissions covers, written down. |
| hooks | Your own commands, run as part of a FutureX run. |
| defaultTier | The module a run starts on when you do not pass --tier. |
| verify | How the automatic checks behave after an edit. --no-verify skips them for one run. |
| mcpServers | MCP servers the CLI connects to. |
Permissions are per project on purpose: consent you gave one repository should never follow you into another one.
Environment
One variable. FX_API_KEY in a .env file is loaded automatically when you run futurex.
FX_API_KEY=your_key_hereNo key on hand? Run futurex login and sign in through the browser with a link code. Then futurex whoami tells you which backend and auth mode you landed on, and futurex logout clears it again.
Either way you are billed by usage: $1 per 1M tokens on Fast, $2 per 1M on Pro, with a free first month on new accounts. The full breakdown is on pricing.